How to Detect AI-Generated Text?

Artificial intelligence has penetrated almost every area of our lives, and content creation is no exception. Tools such as ChatGPT, Gemini, and Claude have seen exponential growth in users over the past two years, and AI-generated texts can now be found on blogs, social media, emails, and perhaps even on the menu at the restaurant where we decide to have dinner. 

Generating text with the help of AI represents an extraordinary opportunity to streamline marketing processes. AI can be extremely cost- and time-efficient. However, it also has its dark sides. In order to achieve quality results and maintain a high professional standard, it is essential to apply the “human-in-the-loop” principle. This concept emphasizes that in order to achieve a truly valuable result, AI-generated output requires human input, supervision, validation, and refinement. 

After discussing how to detect AI-generated images in a previous article, the purpose of today’s article is to guide you in discerning the “marks” of AI in text and thus be able to create better, more informative, and more authentic texts. What we aim for as marketers and, in general, as IT professionals, is the intelligent use of AI tools, combining the efficiency of technology with the professionalism and creativity of a specialist.

Content Marketing in the Era of AI: Principles Remain

Before exploring the distinctive features of AI texts, it is important to remember the basic characteristics of quality content. Good material is based on: 

• rigorous documentation, preferably from multiple reputable sources, including citations that lend credibility
• identification with the target audience, understanding and addressing the context in which it is presented
• compliance with text structuring rules for various purposes
• optimization for both traditional search engines (SEO) and new generative engines (GEO – Generative Engine Optimization).

Why is it important to identify AI texts?

As marketers who work with content on a daily basis, it is important to remain vigilant about the indiscriminate use (copy-paste) of AI-generated texts, for several reasons:

1. The illusion of correctness: AI texts often appear correct, fluent, and well-organized. Perhaps even too correct. Despite these appearances, however, they may contain “hallucinations” (false information presented as real facts), omissions, subtle logical problems, or other oversights that can mislead the reader.
2. Professional credibility: Once an informed reader detects that a text is generated by AI, without human input, your credibility as a specialist or your brand’s credibility may suffer. Authenticity is a valuable currency in marketing and beyond.
3. Indexing and visibility: There is a tendency for search engines, especially those based on LLMs, to index purely AI-generated content less effectively. AI models avoid training on data also created by AI, because this process, which we could call incestuous, leads over time to a degradation in the quality of the models (“model drift”) and to outputs that are less faithful to reality.
4. Creating a long-term strategy: At present, search engine rules and international AI legislation are extremely volatile and changing rapidly, so we cannot predict how AI-generated content will be viewed in the future. Using AI as a tool, not as a replacement for specialists, is a technique for future-proofing your marketing strategy, demonstrating a high level of professionalism and adaptability.

How to Recognize AI-Generated Text

Mass AI models, known as “large language models” (LLMs), use a technique called “next token prediction.” Essentially, this is a highly sophisticated autocomplete function that predicts which word is statistically most likely to follow in a phrase or sentence, based on the huge amount of text data the model has been trained on. This process, although extremely advanced, leaves behind certain stylistic and structural “fingerprints”.

Identifying AI-generated text can be based on several indicators, from subtle technical elements to stylistic peculiarities.

Unicode watermarks

Some models may insert invisible or less common characters into the text, details that are usually overlooked when the user copies and pastes AI-generated content. These may include zero-width spaces or less common punctuation marks, such as em dashes (—) or straight quotation marks (” “).

Elements of discourse structure

This is where the clearest differences between human and artificial text can be seen.

Word predictability

• AI tendency: AI models tend to choose the most statistically probable words, as explained above. Thus, based on the examples given in the video above, in a sentence such as “He ate…”, AI will most likely choose “apple” or “cake”, rather than phrases such as “he swallowed the dumpling”.
• Human tendency: People often make less predictable lexical choices, but ones that make perfect sense in context, giving the text originality and style. A human might write “He ate everyone at the meeting for breakfast”.

Sentence rhythm and length

• AI tendency: Artificially generated text often has a monotonous rhythm, with sentences of similar length and structure, being descriptive but somewhat lifeless.
• Human tendency: Human writing is generally more “bursty”. A 20-word sentence may be followed by one with only 3 words. This variation in rhythm makes the text more dynamic and engaging. A text with human “personality” will include shorter, more colloquial sentences, more expressive word choices, and varied grammatical structures.

Unnatural perfection

AI-generated text often sounds too correct. It lacks the idiosyncrasies of personal style, the little poetic licenses, or even the occasional spelling mistakes that humans make. Researchers at Google, quoted by MIT Technology Review, say that when you find a spelling mistake in a text, it is almost certain that it was written by a human. Of course, we do not encourage writing with mistakes, but absolute perfection can be a red flag.

Excessive use of emojis

Especially in texts intended for advertising campaigns (copy for PPC ads, newsletter texts), AI tends to incorporate numerous symbols, which have already become a trademark and reveal the AI text at first glance.

Excessive capitalization

In Romanian, it is easy to recognize AI-generated text as a fragment containing titles or subheadings (paragraph titles) written with the first letter of each word capitalized. It looks strange, doesn’t it? This trend, indiscriminately borrowed from English, is not only incorrect but also provides a clear sign of text taken without editing from the output of an LLM.

Online AI Content Detection Tools

In addition to manual analysis, there are a number of tools (some even based on AI trained specifically for this purpose) that can help you scan a text and estimate the probability (percentage) that it was generated by AI. Among the best-known such tools are:

• Grammarly: In addition to grammar correction features, it also offers an AI content detector.
• Quillbot: Known for its paraphrasing features, it also includes an originality checker and AI detector.
• Turnitin: Widely used in academia for plagiarism detection, it has also integrated advanced technologies for identifying artificially generated texts.

Conclusion: AI Is a Strategic Partner, Not a Replacement

Artificial intelligence is undoubtedly one of the most powerful tools in a content marketing specialist’s arsenal. Wise professionals do not reject it, but learn to use it in a way that maximizes their efficiency and return on investment (ROI). The key to success lies in mastering the techniques of collaborating with AI, using it to generate ideas, structure content, and automate repetitive tasks, while maintaining control over quality, brand voice, and message authenticity.

At NewTech Academy, we encourage you to practice and deepen these techniques. If you want to take your knowledge to the next level and integrate the latest technologies into your strategy, we invite you to learn more about our Digital Marketing Specialist course. It includes a module dedicated to the use of AI in marketing and will prepare you for the challenges and opportunities of the digital future.